Facial recognition "caught" the wrong person, and he spent 10 days in jail

Yang Jingbiance from Aofei Temple
Quantum Bit Report | Public Account QbitAI

It's just outrageous.

I didn't do anything, but was detained by the police for 10 days for no reason?!

This is a true story that happened in New Jersey, USA .

In a hotel theft case, the suspect drove away.

The police did not check fingerprints or conduct DNA tests, but only relied on the facial recognition system to identify the "criminal" - Nijeer Parks .

△ Source: NJ Advance Media

But the embarrassing thing is that this "criminal" has never been to that place and doesn't even know where the hotel is.

In fact, he doesn't even know how to drive and doesn't even have a driver's license.

When he went to the police station, it was his cousin who drove him there, but he was still detained for 10 days.

The real pot fell from the sky

According to NJ Advance Media, a theft occurred at the Hampton Inn in Woodbridge, New Jersey in January last year.

The suspect stole candy and other snacks from the hotel gift shop and fled to a men's restroom after hotel staff called police.

When police arrived, the man handed over a driver's license identifying himself as Jamal Owens .

The police checked the computer in the patrol car and found that the driver's license was invalid and might have been forged by the suspect. Just as the police were about to arrest him, the man escaped from the hotel entrance and jumped into a Dodge Challenger.

△ Probably this model

According to NJ Advance Media, citing court documents, the suspect also rammed their patrol car. An officer followed the car and later found it abandoned in a parking lot.

Then, the magical things started from the next day.

Investigators from New York and New Jersey used facial recognition software to scan driver's license documents and compare them with photos of suspects stored in police and FBI databases, finding people who "highly matched" the photos .

The software, which relies on billions of social media photos to identify suspects, has been criticized, and for reasons that NJ Advance Media did not disclose, New Jersey has banned its use.

Then, police from both states told their colleagues in Woodbridge: We have a suspect!

The suspect was identified as Parks , and a warrant was issued for his arrest charging him with shoplifting, forgery of a government document, resisting arrest, aggravated assault with a weapon, and leaving the scene of an accident.

Police then checked several addresses in an attempt to arrest Parks but were unable to locate him.

Parks learned about the incident through his grandmother and immediately asked his cousin to drive him to the police station to clarify the mistake.

As a result, as soon as he arrived at the police station, he was handcuffed and interrogated by seven or eight police officers at the same time.

Inside, the police showed him pictures of the damaged Dodge Challenger and kept asking, "Do you know what you did?"

After his explanation failed, he was thrown directly into prison.

Parks revealed he stayed there for 10 days before being released on bail .

△ Photos taken after Parks was arrested

At that time, the prison prosecutor reminded him,

If he pleads guilty, in exchange he will serve six years in prison, and he can be released early when he has served 85% of his sentence, including three years of parole.

But Parks believes that if he went to trial, prosecutors would seek a sentence of 20 years or more based on his previous criminal history .

He was convicted of drug trafficking for six years, but after being released from prison, he decided to turn over a new leaf, found a clerical job, and lived a stable life with his fiancée. But this "face recognition" case broke his peace.

Thus, he began a litigation journey that lasted more than a year.

Parks' attorney, Don, filed a lawsuit against the town, the police department and public officials including Mayor John McCormack, alleging that investigators violated his client's rights by relying on facial recognition software.

A spokesman for the mayor said the town had not seen the civil complaint and could not comment.

Now, a Superior Court judge has begun pressuring the prosecutor's office to produce more evidence in the case than just facial recognition software.

Anyway, it looks like it's not over yet.

Does facial recognition have the final say in catching criminals?

Coincidentally, this is not the first time that facial recognition has arrested the wrong person in the United States. Parks' arrest is not the most serious of such incidents.

In 2014, a financial advisor named Steve Tally was falsely accused of two bank robberies.

The first was on May 14 of that year, and the second was on September 5, 10 days before he was arrested and charged with assaulting a police officer during the second robbery.

After police publicly released surveillance images of the September robbery, three acquaintances of Talley called the police hotline and pointed out the similarities between Talley's appearance and the robber. Even Talley's ex-wife thought it was him after seeing the images provided by police.

But Tully had a good alibi: he was working at a company at the time.

The police ignored it and kept Talley in jail for two months, during which time they beat him.

It was not until two months later that Talley was released after his defense attorney produced surveillance footage of him at work.

Can facial recognition be used as definitive evidence to accuse a person?

In 2009, a National Academy of Sciences paper stated that, aside from DNA testing, no other forensic evidence method can reliably and consistently "prove a link between evidence and a specific individual or source."

Technically speaking, it is impossible to determine the uniqueness of a suspect through facial recognition.

The FBI’s forensic audio, video and image analysis department itself knows that when manual facial comparisons are conducted, the conclusions drawn will ultimately be based on personal opinions.

So is the machine’s algorithm necessarily reliable?

As early as 2008, an FBI report recommended that the image department study the frequency of quantified facial features. However, humans have been working on this for more than 100 years, and there is still no conclusion.

Another problem with facial recognition is that surveillance cameras compress video images, causing patterns such as skin, veins and moles used to distinguish suspects to be removed, damaged or distorted.

Talley was arrested because of a mole on the robber's face in a surveillance video in September.

In fact, due to the different lighting conditions of the captured videos, some artifacts may be mistaken for moles.

Although manual comparison was used to capture Tali, today, facial recognition using AI, which has higher accuracy, still faces many problems.

Algorithms can search millions of faces in seconds, and for images with regular lighting and standardized poses, many face recognition systems have surpassed human accuracy. But there is still no algorithm that can guarantee 100% accuracy.

In terms of professionalism, AI algorithms have poor explainability, and facial recognition lacks empirical data to support forensic opinions.

Error rate and privacy are worrying

Speaking of using facial recognition to catch criminals, I wonder if you will remember the "Five Consecutive Kills" at Jacky Cheung's concert a few years ago. At that time, the police relied on facial recognition to catch five fugitives in a row.

But in comparison, the British police were not so lucky. In the 2017 Champions League, the British police used a facial recognition system, which resulted in a total of 2,470 alerts, equivalent to a suspect being alerted every three seconds, with an error rate of 92%.

In daily life, the privacy issue of facial recognition is more worrying than the error rate issue.

Fearing that his privacy might be leaked by the image collector, Guo Bing, an associate professor at Zhejiang Sci-Tech University, took Hangzhou Wildlife World, which used facial recognition technology, to court, making it the "first facial recognition case" in China.

In the end, Animal World deleted the facial feature information that Guo Bing submitted when he applied for an annual pass.

△ Image source: Hangzhou.com

Facial recognition facial information is personal sensitive information. Once leaked, it will cause great harm to the personal and property safety of individuals.

For this reason, people are becoming more cautious about facial recognition.

Last May, San Francisco banned the government from using facial recognition technology for security and privacy reasons, becoming the first city in the world to ban facial recognition.

Technology companies have become increasingly cautious about facial recognition in recent years. IBM has withdrawn from researching this technology, and Microsoft has also deleted the large facial recognition dataset MS Celeb used for research.

Facial recognition, the original "killer" application of AI, has created many AI star companies, but as it has developed to date, facial recognition is also coming to a crossroads.

On the one hand, its convenience has been fully demonstrated; on the other hand, the accompanying privacy controversies and error rate problems continue to become hot searches as new events.

Should facial recognition be used? Where should facial recognition be used? How should facial recognition be used?

These are becoming problems in the new era.

In addition, I heard that many communities have switched to facial recognition?

Reference links:
https://www.nj.com/middlesex/2020/12/he-spent-10-days-in-jail-after-facial-recognition-software-led-to-the-arrest-of-the-wrong-man-lawsuit-says.html
https://theintercept.com/2016/10/13/how-a-facial-recognition-mismatch-can-ruin-your-life/